The expansion of agricultural land at the cost of pastoral land is the common cause of land degradation in the arid areas of developing countries, especially in Morocco. This study aims to assess and monitor the transformation of pastoral land to agricultural land in the arid environment of the Feija Basin (Southeast of Morocco) and to find the key drivers and the issues resulting from this transformation. Spectral mixture analysis was applied to multi-temporal (1975–2017) and multi-sensor (i.e. Multi-spectral Scanner, Thematic Mapper, and Operational Land Imager) Landsat satellite images, from which land use classifications were derived. The remote sensing data in combination with ground reference data (household level), groundwater and climate statistics were used to validate and explain the derived land use change maps. The results of the spatiotemporal changes in agricultural lands show two patterns of changes, a middle expansion from 1975 to 2007, and a rapid expansion from 2008 to 2017. In addition, the overall accuracy demonstrated a high accuracy of 94.4%. In 1975 and 1984, the agricultural lands in Feija covered 0.17 km2 and 1.32 km2, respectively, compared with 20.10 km2 in 2017. Since the adoption of the Green Morocco Plan in 2008, the number of watermelon farms and wells has increased rapidly in the study area, which induced a piezometric level drawdown. The results show that spectral mixture analysis yields high accuracies for agricultural lands extraction in arid dry lands and accounts for mixed pixels issues. Results of this study can be used by local administrators to prepare an effective environmental management plan of these fragile drylands. The proposed method can be replicated in other regions to analyse land transformation in similar arid conditions.